[CPMD-list] CPMD-3.11.1: Error in test runs

[Axel Kohlmeyer]

On Mon, 22 Jan 2007, Amita Wadehra wrote:

hi,

it looks as if your mail got stuck in the mailing list
for a while. :-(

AW> Hi,
AW> 
AW> I compiled serial executable for CPMD-3.11.1 on a Linux machine using the
AW> following PC-IFC-P4:
AW> 
AW>      IRAT=2
AW>      CFLAGS='-c -O2 -Wall'
AW>      CPP='/lib/cpp -P -C -traditional'
AW>      CPPFLAGS='-D__Linux -D__PGI -DFFT_DEFAULT -DLINUX_IFC'
AW>      FFLAGS='-c -w90 -w95 -O2 -pc32 -tpp7 -unroll -tune pn4 -arch pn4'


unless you found it out yourself already. the problem is here in 
the -pc32 flag. please read the compiler documentation carefully
with this switch you turn the floating point unit into 32-bit mode.
CPMD _requires_ double precision floating point (i.e. 64-bit)
even on a 32-bit addressing platform. those two are completely
independents, anyways. the error you see is exactly a manifestation
of the much reduced accuracy (7 significant digits, vs. 14) in 
single precision floating point when you sum up the density over
all grid points.

cheers,
   axel.



AW>      LFLAGS='-lsvml -Xlinker -rpath=/usr/local/mkl-7.0.1/mkl701/lib/32/ \
AW>              -Xlinker -rpath=/usr/local/intel-8.1.033/lib \
AW>              -L/usr/local/mkl-7.0.1/mkl701/lib/32/ -lmkl_lapack -lmkl_p4 \
AW>              -lguide -Vaxlib'
AW>      FFLAGS_GROMOS=' '
AW>      if [ $debug ]; then
AW>        FC='ifort -g'
AW>        CC='gcc -g'
AW>        LD='ifort -i-static -g'
AW>      else
AW>        FC='ifort '
AW>        CC='gcc '
AW>        LD='ifort -i-static '
AW>      fi
AW> 
AW> There were no errors in compilation. However, when I tried to run CPMD
AW> with the inputs given in CMPD-test, it exited with the following error in
AW> the output:
AW> 
AW>  IN FOURIER SPACE:                                34.000003814697
AW>  IN REAL SPACE:                                   33.983016967773
AW>  STOPGM! STACK OF MAIN CALLS:
AW>  STOPGM! CALL    FORCEDR
AW>  STOPGM! CALL     FORCES
AW>  STOPGM! CALL     RHOOFR
AW> 
AW> 
AW>  PROGRAM STOPS IN SUBROUTINE RHOOFR| TOTAL DENSITY SUMS ARE NOT EQUAL
AW> 
AW> This is happening with other inputs that I am using from CPMD-test too. I
AW> have not made any changes in the input file or pseudopotential, so I don't
AW> know where this error is coming from. I read similar queries on CMPD
AW> mailing and one of the replies suggested that it may be caused by
AW> miscompilation. Please let me know if there is something wrong with the
AW> PC-IFC-P4 file that I am using or is there any other source of this error.
AW> 
AW> Thanks
AW> Amita
AW> _______________________________________________
AW> CPMD-list mailing list
AW> CPMD-list at cpmd.org
AW> http://cpmd.org/mailman/listinfo/cpmd-list
AW> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.